Standard risk measures evaluate a portfolio at a single future time horizon. But real risk involves multiple horizons — the same position may be held for a day, a week, or a month, and the risk assessment should account for the possibility that the relevant horizon is itself uncertain.
The paper builds risk measures that incorporate horizon risk alongside interest rate uncertainty. The tool is backward stochastic differential equations (BSDEs), which define the risk measure dynamically: the risk at time t depends on the future evolution, which is itself uncertain.
Note (2026-11-13; read against the paper's full text): the essay has the direction of cash subadditivity wrong at one point. In the paper a risk measure is cash subadditive when rho(X + m) is at least rho(X) - m for m at least 0, that is, adding cash m reduces the risk by at most m, never by more; the essay says the reduction is more than the amount at shorter horizons. The paper also defines horizon risk as using a risk measure designed for long-term positions to evaluate short-term investments, and vice versa; the essay describes it as the relevant horizon being itself uncertain. The paper's results are for fully dynamic cash subadditive measures from BSDEs (Lipschitz and quadratic drivers) and shortfall approaches, with a dual representation for the new h-generalized shortfall measures.
Cash subadditivity replaces cash additivity. A cash-additive risk measure says: adding $1 of cash reduces risk by exactly $1. This assumes a known risk-free rate. With uncertain interest rates, adding $1 of cash reduces risk by an amount that depends on which horizon materializes and what rates prevail — generally less than $1 at longer horizons and more at shorter ones.
The hq-entropic risk measure demonstrates the framework's content. It belongs to the family of generalized shortfall risk measures (quantifying the expected cost of exceeding a threshold) but is not a certainty equivalent (not derivable from a utility function). Classical entropic risk measures are all three — BSDE, shortfall, and certainty equivalent — but the horizon-risk generalization breaks the equivalence. The three characterizations that agree in the single-horizon case diverge when the horizon is uncertain.
Risk at an uncertain horizon is not the same as risk at a fixed horizon with uncertain returns. The horizon is a separate source of risk, and its interaction with interest rate uncertainty creates new mathematical structure.